Sujay Jadhav

Active learning and data cleaning for Object detection - Integration of my master's thesis work on Bosch production dataset in Automated driving - MLOps, Microsoft Azure, AML, SQL, MLFlow

Research on dataset pruning and active learning strategies for object detection - Data Strategy team - machine learning - Traffic sign recognition - Data Cleaning, pruning and feature engineeing

Research project on data strategy for video perception in autonomous driving. • PoC strategy to come up with plausible 'data-quality' metrics using the training dynamics of gradient-based models (predominantly CNN variants and visual transformer). Tried it on standard torchvision (MNIST, CIFAR100) as well as detection (GTSRB) datasets. • Exploring the application of these metrics in dataset pruning and instance selection for active learning, among other applications

• Developing/evaluating custom-CNN, for switchgear's health monitoring algorithm with IR (infrared image) data • Transfer learning (VGG16, EfficientNetB0) with lambda layers, subclassing • Synthetic over-sampling for class-imbalance, and Evaluation Metrics w/ F1, F-beta, Confusion matrix, G-mean Sensitivity, ROC curve • Integrated state-of-the-art EDSR and SRGAN algorithms for infrared-image superresolution, in ABB Global Hackathon 2021 • Research on One-Shot learning, Multi-Hot encoding, LIME

• Developed Image/Map outline extraction libraries for precision farming using dummy data over CAN 2.0, ISOBUS in Off-road vehicles • MATLAB-Simulink Modelling for Hardware-in-Loop regression Tests • Updated firmware code for OCR (Over-current Relay) in C++ • Featured database formation for blob-detection in faulty-contactor images using Histogram of Oriented-Gradients (HoG) • 'Star-of-the-Month' reward for optimizing Text comparison (Optical Char. Recognition) libraries used throughout the team

● Specialization: -BILDVERSTEHEN (Image Understanding and Computer Vision) -MASCHINELLES LERNEN (Neurocomputing) -KÜNSTLICHE INTELLIGENZ (Probability & Learning Theory) ● Hauptseminar: Depth Estimator- monocular Camera using CNN ● Technical Report: 'Predictive maintenance' using TPMS (Tire Pressure Monitoring System)

8.53/10 CGPA - [180 ECTS] ● Seminar: 'Neighborhood Operators' and 'Edge detection' in Image Processing (Matlab) ● Semester-Internship: R&D Johari Digital Healthcare India Ltd. (Biomedical product design and documentation) ● Semester exchange at Hof University, Germany ● Programming: Basics of C, OOP, Python, VHDL, Matlab-Simulink